|
Sébastien Arnaud
arnaudsj at mac.com
Mon Aug 9 21:38:41 EDT 2004
Grisha, Thank you for your enlightenment in those dark areas of my assumptions (oh man, let's say it, I was wrong on the whole HTTP 1.1 thing ;) Here are 2 simple tests that hopefully will speak for themselves better than I can explain it. I ran with apache bench (2.40) against one simple test page rendered by mod_python. ############# BEGIN: FIRST TEST ############## ab -v 3 -n 1 -k http://localhost/test/index.py This is ApacheBench, Version 2.0.40-dev <$Revision: 1.121.2.10 $> apache-2.0 Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/ Copyright (c) 1998-2002 The Apache Software Foundation, http://www.apache.org/ Benchmarking localhost (be patient)...INFO: POST header == --- GET /test/index.py HTTP/1.0 User-Agent: ApacheBench/2.0.40-dev Connection: Keep-Alive Host: localhost Accept: */* --- LOG: header received: HTTP/1.1 200 OK Date: Tue, 10 Aug 2004 01:19:55 GMT Server: Apache Cache-Control: no-cache="set-cookie" Set-Cookie: testcookie=testvalue2; expires=Tue, 10-Aug-2004 01:24:56 GMT Connection: close Content-Type: text/html <html> <head> <meta http-equiv="Content-Type" content="text/html; charset=UTF-8"> <title>Test XSLT View</title> </head> <body><p>The message is: <span></span></p></body> </html> LOG: Response code = 200 ..done Server Software: Apache Server Hostname: localhost Server Port: 80 Document Path: /test/index.py Document Length: 178 bytes Concurrency Level: 1 Time taken for tests: 0.17957 seconds Complete requests: 1 Failed requests: 0 Write errors: 0 Keep-Alive requests: 0 Total transferred: 406 bytes HTML transferred: 178 bytes Requests per second: 55.69 [#/sec] (mean) Time per request: 17.957 [ms] (mean) Time per request: 17.957 [ms] (mean, across all concurrent requests) Transfer rate: 0.00 [Kbytes/sec] received Connection Times (ms) min mean[+/-sd] median max Connect: 0 0 0.0 0 0 Processing: 17 17 0.0 17 17 Waiting: 8 8 0.0 8 8 Total: 17 17 0.0 17 17 ################# END FIRST TEST ##################### Note, the "Connection: close" returned by apache even though ab requested a "Connection: Keep-Alive" Now, the same request, but the simple test.py script is now setting req.headers_out['Content-Length'] ################ BEGIN SECOND TEST ################### ab -v 3 -n 1 -k http://localhost/test/index.py This is ApacheBench, Version 2.0.40-dev <$Revision: 1.121.2.10 $> apache-2.0 Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/ Copyright (c) 1998-2002 The Apache Software Foundation, http://www.apache.org/ Benchmarking localhost (be patient)...INFO: POST header == --- GET /test/index.py HTTP/1.0 User-Agent: ApacheBench/2.0.40-dev Connection: Keep-Alive Host: localhost Accept: */* --- LOG: header received: HTTP/1.1 200 OK Date: Tue, 10 Aug 2004 01:19:03 GMT Server: Apache Cache-Control: no-cache="set-cookie" Set-Cookie: testcookie=testvalue2; expires=Tue, 10-Aug-2004 01:24:03 GMT Content-Length: 178 Keep-Alive: timeout=30, max=1000 Connection: Keep-Alive Content-Type: text/html <html> <head> <meta http-equiv="Content-Type" content="text/html; charset=UTF-8"> <title>Test XSLT View</title> </head> <body><p>The message is: <span></span></p></body> </html> LOG: Response code = 200 ..done Server Software: Apache Server Hostname: localhost Server Port: 80 Document Path: /test/index.py Document Length: 178 bytes Concurrency Level: 1 Time taken for tests: 0.282956 seconds Complete requests: 1 Failed requests: 0 Write errors: 0 Keep-Alive requests: 1 Total transferred: 466 bytes HTML transferred: 178 bytes Requests per second: 3.53 [#/sec] (mean) Time per request: 282.956 [ms] (mean) Time per request: 282.956 [ms] (mean, across all concurrent requests) Transfer rate: 0.00 [Kbytes/sec] received Connection Times (ms) min mean[+/-sd] median max Connect: 0 0 0.0 0 0 Processing: 282 282 0.0 282 282 Waiting: 247 247 0.0 247 247 Total: 282 282 0.0 282 282 #################### END SECOND TEST ############################# Note, that suddenly here Apache replies properly to the keep-alive request with the response headers: "Keep-Alive: timeout=30, max=1000" and "Connection: Keep-Alive". All of this because we did set the "content-length" header in the python code. Anybody, who has a more in depth knowledge of Apache, tempted to solve this mystery? Cheers, Sébastien On Aug 9, 2004, at 9:12 AM, Gregory (Grisha) Trubetskoy wrote: > > > On Sun, 8 Aug 2004, [ISO-8859-1] S?bastien Arnaud wrote: > >> First, I was wrong when I said mod_python was forcing the HTTP Header >> "Connection: Close", it is in fact responding to a client request >> http keep-alive, but it is missing an important part: the HTTP header >> "Content-Length: xxx"! > > If this is HTTP 1.1, then absense of content-length indicates that the > browser wants chunked-encoding, which means that the output is a > number indicating size, followed by a chunk of data. With HTTP 1.1 > content-length is obsolete. There is a pretty good description of it > here: > > http://www.apacheweek.com/features/http11 > >> Since it is the only way browsers can keep connections alive (to know >> how much bytes they need to expect back from the server), > > ...not with HTTP 1.1 > >> without it, most browsers will NOT keep the connection alive (maybe >> your version of Firefox is sensitive to this missing header and tries >> to wait for more content to come back from the server) > > No, this doesn't make a lot of sense. :-) The browser isn't keeping it > alive because it is seeing a Connection: close, most likely. > >> For the record I got mistaken in my early tests because I used Lynx >> to get the HTTP headers back, and found out that Lynx did not send a >> valid HTTP 1.1 request with keep-alive (duh!), therefore it was ok to >> receive a "Connection: Close" back then. >> >> Anyway, I think it would be nice to have an explanation why >> mod_python is not setting by default the Content-Length HTTP header >> when it is already setting the "Connection: Keep-alive" HTTP header >> in response to a keep-alive request from the client. > > Mod_python does not set the connection header anywhere in the code, > but other code in Apache probably does. In my experience it is very > smart about it, so there is more than likely a good reason for it. > Something is telling Apache to send a "close"... It could the > configuration, some module in Apache, and it could be something that > the browser has caused (without seeing the actual headers its hard to > tell). > > Grisha > _______________________________________________ > Mod_python mailing list > Mod_python at modpython.org > http://mailman.modpython.org/mailman/listinfo/mod_python -------------- next part -------------- A non-text attachment was scrubbed... Name: PGP.sig Type: application/pgp-signature Size: 155 bytes Desc: This is a digitally signed message part Url : http://modpython.org/pipermail/mod_python/attachments/20040809/2ec59bcf/PGP.bin
|